I'm a student. And I'm very bothered to see how my fellow students are using LLMs. It worries me, as it's ruining the classes students spend hours on. I'll explain what I notice my fellow students do each day. This is a rant, but I hope it could also serve as a wake-up call to some fellow students, or maybe a teacher.
A concrete example that I witnessed just yesterday: the teacher asks us to fill in something called a 'core competence tree' for a business. Each student got assigned a different company, we got 15 minutes and then some students would be chosen to present the thing.
A core competence tree is a strange thing. If you didn't pay attention in class, you'd have a hard time making one. Some of my classmates also got assigned some unknown company. The person in front of me had to make an analysis of Cling Datacenters. You've guessed it: he fires up ChatGPT, types in 'make core competence tree of cling datacenters' and hits enter. This is the output when I try to reproduce it:
Creating a core competence tree for Cling Datacenters involves
identifying the key competencies that make the company
unique and successful. Here's a conceptual outline of what this
might look like:
### Core Competence Tree for Cling Datacenters
#### Roots (Core Business)
- **Data Center Operations**: Efficient and reliable management of data center facilities.
- **Customer Service**: Providing exceptional support and service to clients.
#### Trunk (Core Products)
- **Cloud Services**: Offering scalable and secure cloud solutions.
- **Colocation Services**: Providing space, power, and cooling for clients' IT equipment.
#### Branches (Core Competencies)
- **Energy Efficiency**: Implementing green technologies and practices to reduce energy consumption.
- **Security**: Ensuring the highest levels of physical and cyber security.
- **Scalability**: Ability to scale services according to client needs.
- **Innovation**: Continuously improving and innovating data center technologies.
#### Leaves (End Products)
- **Managed Services**: Comprehensive IT management and support.
- **Disaster Recovery**: Solutions for data backup and recovery.
- **Network Services**: High-speed and reliable network connectivity.
This tree structure helps visualize how the core business
and products support the competencies that lead to the end
products and services offered by Cling Datacenters.
If you need a more detailed or specific model, feel free
to let me know!
He typed the output in a nice graph and called it a day.
Now, I'm hoping I don't have to explain why this is a bad idea. Just for the sake of clarity, though, I will:
- Student learns nothing
- ChatGPT has never heard of Cling Datacenters
- ChatGPT doesn't know how to properly make a core competence tree
- ChatGPT made all of it up
Get an assignment? Put the whole thing in ChatGPT. Need to write up a list of things to consider in terms of sustainability when making an investment? Open ChatGPT. Will something take more than 5 seconds of thinking? Chat. Didn't pay attention in class? Ask Chat.
All of my classmates assume whatever the LLM spits out is correct. Without an exception. Let me quote someone I worked with on a project recently: "ChatGPT said it so it must be right".
I honestly don't know anymore what LLMs are even good for. Getting inspiration, maybe? Asking questions that you then have to do research on yourself?
But: in no case can you expect ChatGPT to do the learning for you. It's not possible. Yet, that is literally what students do, all the time. The assignments which are supposed to make you learn a certain theory are outsourced to an LLM. The output is accepted and turned in without the student taking a second look. Result: the pupil turned in a worthless assignment, without ever thinking about the assignment himself.
Solution? Not sure. Probably involves educating the students about generative AI, showing practical examples of the tool failing at certain tasks. Helping them understand when it fails and why.
It'll be alright. There's no way for students to use ChatGPT on their exams. At least not on my university. The students know this, and thus practice for tests. I don't think a whole generation will go down. But if LLMs are allowed to be used for assignments in class, and students aren't made aware of the limits of these tools, classes have just become a whole lot less effective.
I see my fellow students using ai to come up with the thesis for their essays, which is crazy. It's such a high school mentality. The message I would want to send to every student is that they are better than ai.